InfoScope boundary
CA-A1 architecture record. Updated: 2026-10-05. Architecture boundary only — does not activate CA-AI1 or corpus ingestion.
Permitted role (advisory surfaces)
Future Cicada InfoScope may assist with:
- research retrieval and corpus search;
- evidence summaries;
- precedent comparison across cases;
- research-question generation;
- evidence discovery (surfacing candidate sources for human review);
- advisory suggestions to operators.
Every one of these produces artefacts labelled as AI-derived, carrying provenance that records the model/process that produced them.
Prohibited role (hard boundary — structural, not policy)
InfoScope must not:
- select participants;
- choose or alter strata, quotas or weighting;
- rank people;
- approve evidence;
- define or hold institutional authority;
- make assembly decisions;
- turn suggestions into canonical decisions;
- alter closed recommendations;
- automatically write into authoritative lifecycle records.
How the architecture enforces it
ai_artefactflag — any record derived from AI/InfoScope output is structurally marked; it can never carry authority-class status.- Advisory ≠ authoritative class — AI-derived material can attach only to non-authoritative fields (evidence discovery suggestions, research notes); object schemas for consequential states require an
authority_ref, which an AI process can never hold. - No write path to canonical records — ingestion flows produce proposals reviewed by humans; the canonical write is a human-attributed event.
- Distinguishability — published surfaces must render AI-derived material visibly distinct from canonical institutional records.
- Digest discipline — AI artefacts cited by canonical records are digest-pinned so later drift is detectable.
Boundary to corpus itself
This architecture record concerns runtime assistance. Use of this research corpus by InfoScope for retrieval is the CA-AI1 contract — separately gated, untouched here.